We have added experimental support for GPU passthrough in Origo OS. Specifically we initially support Nvidia and AMD GPUs. When starting VMs or building Stacks, you may specify how many GPUs your VM requires and the minimum amount of VRAM. The scheduler in Origo OS will then try to locate a compute node with the required ressources. If successful, the VM starts, if not, launching the VM fails. GPU passthrough and subsequent release of the GPU is handled by Origo OS with a few caveats.
GPU and vRAM usage and the associated metered billing has been built into Origo OS and is shown with the usual usage and billing information. If you use Origo Public Cloud Toolkit, the UI, billing and ressource handling has also been updated to take into account GPUs.
Here is a screen-shot from Origo Public Cloud Toolkit that shows launching a stack which asks for a GPU (prices are just examples, you can of course charge what you want).
